I read an interesting article on The Critical Role of Automation in a Virtualized World. The author talks about how all the increases in efficiency that come with virtualization bring a vast increase in complexity that threatens mission-critical business processing. He talks about enterprise schedulers with automation capabilities that provide a solution.

Networking just exacerbates the issues raised in that article. In addition to bringing up the VM on the appropriate ESX server when it is needed, you also have to configure the VLAN to connect it to the right place, and put the firewall ACL in place to allow other servers to acess it. In case you are doing this balancing act across multiple data centers, you also have to configure VPN tunnels and potentially routes to allow the applications to work.

The benefit to all this added complexity, of course, is an even greater increase in the efficiency that you set out to achieve. If you can increase efficiency by flexibly moving VMs across racks, think about how much better it would be if you could move VMs to any location on the globe where the power happened to be cheapest at that minute. Of course, with the cost and latency of moving the bits across the network, its not as simple as that, but you can think about it as increasing the solution space that your enterprise scheduling tools have to play with in order to find the cheapest solution that satisfies your QoS and reliability constraints.